Set up and functions

Analyse choice data

We observed a significant recipient*effort*reward interaction (odds ratio (OR)=0.81, 95% confidence interval=[0.69, 0.95], p=0.009), showing that people were less willing to help others at higher effort and lower reward levels. We also observed significant interactions between recipient and reward (OR=1.71 [1.44, 2.04], p=0.001), effort and reward (OR=1.23 [1.11, 1.36], p=0.001), and main effects of recipient, effort, and reward (Figures 2a-b, S2 and Table S1).

Analyse Κ parameters

This analysis showed that discount values for other were significantly higher for other (Κother median=0.15) than for self (Κself median=0.07, Wilcoxon two-sided signed rank test Z=-5.34, effect size r=0.50, [0.31, 0.65], p<0.001, Figure 2f).

Analyse force data

We observed a significant 3-way interaction between effort, reward, and recipient (χ2(16)=39.02, p=0.002). We also found significant interactions between recipient and reward (χ2(4)=12.49, p=0.022), effort and reward (χ2(16)=47.78, p=0.001), and main effects of recipient, effort, and reward (all χ2s>24.74, all ps<0.001; Figure 2g-h and Table S2). Importantly, there was no significant difference in success between self (mean=0.98) and other trials (mean=0.98, Wilcoxon two-sided signed rank test Z=-0.59, effect size r=0.04, [0.00, 0.28], p=0.55) and Bayesian evidence for no difference (BF01=4.35, substantial evidence in support of the null).

Figure 2

Figure 2. Choice, force, and computational modelling of prosocial and self-benefitting decisions. (a) Participants were less willing to accept the work offer over the rest offer as the effort level increased (GLMM odds ratio for effort (OR)=0.24, 95% confidence interval=[0.22, 0.27], p=0.001). (b) The proportion of work offers accepted over the baseline option increased as reward increased (reward OR=2.54 [2.31, 2.79], p=0.001). Across effort and reward levels, participants were less willing to work when the other person would benefit than when they would benefit (recipient OR=9.37 [7.80, 11.26], p=0.001). This tendency to work more for self than others was most pronounced at the higher reward levels (recipient*reward OR=1.71 [1.44, 2.04], p=0.001) and particularly when a high level of effort was required (recipient*effort*reward OR=0.81 [0.69, 0.95], p=0.009). Data are represented as mean ± SE. (c) We compared a range of computational models and effort discounting that varied in terms of whether the model had a single or separate discount (Κ) parameter(s) for Self and Other trials (models 1-6 vs. models 7-12) and whether the shape of the discount function was parabolic (models 1, 4, 7,10) linear (models 2, 5, 8, 11) or hyperbolic (models 3, 6, 9, 12). Models 7 and 10 had the lowest Bayesian Information Criterion (BIC) scores. These were both parabolic and had separate Κ parameters for self and other. However, model 7 that contained a single choice stochasticity parameter (β) explained behaviour in the majority of participants and was selected as the winning model. Bars show model BIC, proportions show the number of participants with the lowest BIC. (d) Equation for the winning parabolic model with separate discount (Κ) parameters and a single choice stochasticity (β) parameter that explained behaviour in the majority of participants. (e) Parameter recovery using simulated data from the winning model and choice schedule showed excellent recovery of the model parameters. (f) Statistical comparison of the Κ parameters from the best fitting model showed that participants had a lower Κ parameter for self-benefitting compared to prosocial choices (Wilcoxon two-sided signed rank test Z=-5.34, r(38)=0.50, [0.31, 0.65], p<0.001). Data are represented as median ± SE, *** shows p<0.001 in Wilcoxon two-sided signed rank test. (g) Force exerted (normalised areas under the curve during the effort period) for each of the levels of effort. Participants exerted less force for others overall (LMM effect of recipient χ2(1)= 24.74, p=0.001) and there was a 3-way interaction between recipient, effort and reward. (h) Force exerted for each reward level shows participants exerted more force for higher rewards (reward χ2(4)= 66.80, p=0.001) but this effect was reduced when the other person would benefit (recipient*reward χ2(4)= 12.49, p=0.022).

RSA analysis

Multivariate patterns in all of our four ROIs showed significant correlations with the other-effort RDM when making prosocial choices (other-effort mean rank correlation τA ± SE: ACCg = 0.026 ± 0.009, p=0.005; TPJ = 0.033 ± 0.010, p=0.001; AI = 0.021 ± 0.008, p=0.006; dACC/dmPFC = 0.029 ± 0.008, p=0.001, surviving FDR correction for 24 comparisons (6 models, 4 brain areas, 2 recipients)). In contrast for self-effort patterns, only the TPJ brain RDM significantly correlated with the effort model RDM (self-effort mean rank correlation τA ± SE: ACCg = 0.002 ± 0.009, p=0.61; TPJ = 0.024 ± 0.010, p=0.026; AI = 0.008 ± 0.009, p=0.40; dACC/dmPFC = 0.016 ± 0.010, p=0.16). Critically, although TPJ, AI and dACC/dmPFC also represented prosocial effort, they did not do so more strongly than for the self-effort RDMs (Wilcoxon two-sided signed rank test, all ps>0.07). In contrast, the ACCg ROI carried a multivariate representation of effort on prosocial trials only, being the only ROI to display a significant difference between the other-effort and self-effort RDMs (Wilcoxon two-sided signed rank test Z=-2.73, r(38)=0.44, [0.13, 0.69], p=0.006, Figure 3a). Notably the specificity for others effort was not due to overall activation differences for other and self representations, as ACCg was the only ROI that represented other and self offers as equally dissimilar (see Methods, Bayesian paired sample t-test BF01=4.61, substantial evidence in support of the null). Moreover, whilst patterns in several regions significantly correlated with the other-reward RDM (other-reward mean rank correlation τA ± SE: ACCg = 0.009 ± 0.007, p=0.16; TPJ = 0.016 ± 0.007, p=0.027; AI = 0.020 ± 0.008, p=0.006; dACC/dmPFC = 0.025 ± 0.007, p=0.001), no region significantly represented others’ rewards more strongly than self-rewards (Table S3). Thus, multivariate patterns in the ACCg represented effort costs specifically when making prosocial but not self-benefitting choices (see Table S4 for exploratory whole-brain searchlight results).



None of our ROIs showed a significantly stronger correlation of the self-effort than other-effort RDM, nor for the self-SV rather than other-SV RDM (all zvals= <1.37 || >1.56, all p’s>0.12, see Table S3 for reward RDM results).



We found significant correlations between the self-SV and other-SV RDMs in both the dACC/dmPFC and AI (self-SV mean rank correlation τA ± SE: dACC/dmPFC = 0.063 ± 0.012, p<0.001; AI = 0.047 ± 0.012, p<0.001; other-SV mean rank correlation τA ± SE: dACC/dmPFC = 0.073 ± 0.012, p<0.001; AI = 0.051 ± 0.013, p<0.001; all survive FDR correction, Figure 4c).



Outside of dACC/dmPFC and AI, we found multivariate patterns of subjective value that overlapped between self and other in TPJ and ACCg self-SV mean rank correlation τA ± SE: TPJ = 0.026 ± 0.011, p=0.018; ACCg = 0.055 ± 0.012, p<0.001; other-SV mean rank correlation τA ± SE: TPJ = 0.044 ± 0.011, p<0.001; ACCg = 0.038 ± 0.014, p=0.009; all survive FDR correction).

Analyse QCAE

We found that affective empathy was positively correlated with the strength of prosocial effort patterns in ACCg (Pearson’s r(38)= 0.39, p=0.017) whilst cognitive empathy was not (Pearson’s r(38)= 0.05, p=0.78, correlations significantly different t=2.04, p=0.02). For the tracking of prosocial effort in ACCg during force exerted, neither affective or cognitive empathy were significantly correlated (all rs> -0.18, all ps>0.29).

Figure 3

Figure 3. ACCg codes patterns of effort for others only, varies with level of affective empathy, and tracks force required to benefit others only. (a) Across an independent structural ROI of the anterior cingulate gyrus (Neubert et al., 2014), representational dissimilarity patterns of effort were encoded specifically for others. Kendall’s τA shows a greater correlation between brain RDM and effort RDM during the offer period for other than self (Wilcoxon two-sided signed rank test Z=-2.73, r(38)=0.44, [0.13, 0.69], p=0.006. ROI displayed on an anatomical scan of the medial surface. Variability in ACCg effort patterns for others was explained by individual difference in affective empathy, as measured by the Questionnaire for Cognitive and Affective Empathy (QCAE; Reniers et al., 2011). In contrast, there was no significant correlation with cognitive empathy and the two corelations were significantly different from one another (t=2.04, p=0.024). (b) A cluster in the ACCg tracked with the amount of force required during the force period specifically when making prosocial but not self-benefitting decisions (x=-6, y=24, z=20, Z=3.28, k=41, p<0.05, FWE-SVC). Activation overlaid on an anatomical scan of the medial surface. *p<0.05, **p<0.01 ***p<0.001.

Figure 4

Figure 4. Self-benefitting and domain general representations and tracking of subjective value. (a) A cluster putatively in the ventral tegmental area (VTA) encoded representational patterns of subjective value exclusively on self-benefitting trials (x=4, y =-22, z=16, k=291, Z=4.45, p=0.033 FWE-whole brain corrected after thresholding at p<0.001, Figure 4a, Table S4). (b) A sub-region of the ventral anterior insula (vAI; x=-44 y=10 z=-10, Z=3.72, k=59, p<0.05 FWE-small volume) tracked subjective value of the chosen offer trial-by-trial more strongly for self-benefitting than other-benefitting choices. (c) The dACC/dmPFC and AI showed significant correlations between the brain-RDM and subjective value RDM pattern for both other and self offers, consistent with a domain general response in these regions (dACC/dmPFC: τA = 0.064 ± 0.017, p<0.001; AI: τA = 0.047 ± 0.012, p<0.001). (d) Univariate analysis also showed trial-by-trial tracking of subjective value in dACC/dmPFC and AI for both self and other (x=8 y=26 z=34, Z=4.75, k=1033, p<0.05 FWE-whole brain) and bilateral anterior insula (left: x=-28 y=22 z=6, Z=4.47, k=306, p<0.001 FWE-SVC; right: x=34, y=24, z=2, Z=4.38, k=222, p<0.05 FWE-small volume). *p<0.05, **p<0.01 ***p<0.001.

Supplementary

Figure S3

Figure S3. Model identifiability. Data simulated from each of the 12 models for 100 participants shows that the model comparison procedure identifies the model that simulated the data, demonstrated by the strong diagonal. (a) We repeated the simulations and fittings ten times and quantified the winning model as with the modelling of participants’ data as the model with the lowest Bayesian Information Criterion (BIC), summing the number of times that model won across the ten runs. (b) We also calculated the percentage of simulated participants for which each model had the best fit to the data and averaged this over the ten runs.

Table S1

Generalised linear mixed-effects model predicting choices
Parameter Coefficient SE CI CI_low CI_high z Chisq df p
(Intercept) 1.38 0.36 0.95 0.82 2.32 1.23 NA NA NA
Recipient (Self vs. Other) 9.37 0.88 0.95 7.80 11.26 23.92 864.24 1 0.001
Effort 0.24 0.01 0.95 0.22 0.27 -26.72 1702.03 1 0.001
Reward 2.54 0.12 0.95 2.31 2.79 19.29 1035.33 1 0.001
Recipient (Self vs. Other) * Effort 0.93 0.08 0.95 0.79 1.10 -0.86 0.05 1 0.85
Recipient (Self vs. Other) * Reward 1.71 0.15 0.95 1.44 2.04 6.07 33.64 1 0.001
Effort * Reward 1.23 0.06 0.95 1.11 1.36 4.13 10.38 1 0.001
Recipient (Self vs. Other) * Effort * Reward 0.81 0.07 0.95 0.69 0.95 -2.63 6.95 1 0.009

Table S2

Linear mixed-effects model predicting normalised force
Parameter Chisq df p
Recipient (Self vs. Other) 24.7 1 0.001
Effort 6995.3 4 0.001
Reward 66.8 4 0.001
Recipient (Self vs. Other) * Effort 7.9 4 0.11
Recipient (Self vs. Other) * Reward 12.5 4 0.022
Effort * Reward 47.8 16 0.001
Recipient (Self vs. Other) * Effort * Reward 39.0 16 0.002

Table S3

Kendall’s τA correlations between brain RDMs and model RDMs
Recipient Measure Area mean sem pval pvalFDR
Other effort RDM ACCg 0.03 0.01 0.005 0.009
Other effort RDM AI 0.02 0.01 0.006 0.009
Other effort RDM dACC 0.03 0.01 0.001 0.003
Other effort RDM TPJ 0.03 0.01 0.001 0.003
Self effort RDM ACCg 0.00 0.01 0.61 0.61
Self effort RDM AI 0.01 0.01 0.40 0.42
Self effort RDM dACC 0.02 0.01 0.16 0.17
Self effort RDM TPJ 0.02 0.01 0.026 0.032
Other reward RDM ACCg 0.01 0.01 0.16 0.17
Other reward RDM AI 0.02 0.01 0.006 0.009
Other reward RDM dACC 0.03 0.01 0.001 0.003
Other reward RDM TPJ 0.02 0.01 0.027 0.033
Self reward RDM ACCg 0.04 0.01 <0.001 <0.001
Self reward RDM AI 0.03 0.01 <0.001 0.002
Self reward RDM dACC 0.04 0.01 <0.001 <0.001
Self reward RDM TPJ 0.03 0.01 0.005 0.008
Other subjective value RDM ACCg 0.04 0.01 0.009 0.013
Other subjective value RDM AI 0.05 0.01 <0.001 <0.001
Other subjective value RDM dACC 0.07 0.01 <0.001 <0.001
Other subjective value RDM TPJ 0.04 0.01 <0.001 <0.001
Self subjective value RDM ACCg 0.05 0.01 <0.001 <0.001
Self subjective value RDM AI 0.05 0.01 <0.001 <0.001
Self subjective value RDM dACC 0.06 0.01 <0.001 <0.001
Self subjective value RDM TPJ 0.03 0.01 0.018 0.024